U.S. patent application number 13/598268 was filed with the patent office on 2013-02-21 for rolling shutter reduction based on motion sensors.
This patent application is currently assigned to APPLE INC.. The applicant listed for this patent is Jianping Zhou. Invention is credited to Jianping Zhou.
Application Number | 20130044230 13/598268 |
Document ID | / |
Family ID | 47712398 |
Filed Date | 2013-02-21 |
United States Patent
Application |
20130044230 |
Kind Code |
A1 |
Zhou; Jianping |
February 21, 2013 |
ROLLING SHUTTER REDUCTION BASED ON MOTION SENSORS
Abstract
This disclosure pertains to devices, methods, and computer
readable media for reducing rolling shutter distortion effects in
captured video frames based on timestamped positional information
obtained from positional sensors in communication with an image
capture device. In general, rolling shutter reduction techniques
are described for generating and applying image segment-specific
perspective transforms to already-captured segments of a single
image or images in a video sequence, to compensate for unwanted
distortions that occurred during the read out of the image sensor.
Such distortions may be due to, for example, the use of CMOS
sensors combined with the movement of the image capture device. In
contrast to the prior art, rolling shutter reduction techniques
described herein may be applied to captured images or videos in
real-time or near real-time using positional sensor information and
without intensive image processing that would require an analysis
of the content of the underlying image data.
Inventors: |
Zhou; Jianping; (Fremont,
CA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Zhou; Jianping |
Fremont |
CA |
US |
|
|
Assignee: |
APPLE INC.
Cupertino
CA
|
Family ID: |
47712398 |
Appl. No.: |
13/598268 |
Filed: |
August 29, 2012 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
13209899 |
Aug 15, 2011 |
|
|
|
13598268 |
|
|
|
|
13209899 |
Aug 15, 2011 |
|
|
|
13209899 |
|
|
|
|
Current U.S.
Class: |
348/208.6 ;
348/E5.046; 382/275 |
Current CPC
Class: |
H04N 5/3532 20130101;
H04N 5/2329 20130101; H04N 5/23258 20130101; H04N 5/23267
20130101 |
Class at
Publication: |
348/208.6 ;
382/275; 348/E05.046 |
International
Class: |
G06K 9/40 20060101
G06K009/40; H04N 5/228 20060101 H04N005/228 |
Claims
1. A method for reducing rolling shutter distortion in a frame of
image data, comprising: estimating motion of a base row of the
frame based on motion sensor data representing motion of image
sensor during capture; for a plurality of other rows of the frame,
estimating a motion difference between motion of each other row and
the base row based on the motion sensor data; generating a
perspective transform for each other row based on the estimated
motion difference; and applying the generated perspective
transforms to image data of the rows to generate corrected frame
data.
2. The method of claim 1, wherein the base row is a center row of
the image.
3. The method of claim 1, wherein the base row is a center row
located in an identified object of the image.
4. The method of claim 1, wherein estimating the motion difference,
generating the perspective transformation and applying the
generated perspective transform is performed on each row of the
frame.
5. The method of claim 1, wherein estimating the motion difference,
generating the perspective transformation and applying the
generated perspective transform is performed on each anchor row of
the frame.
6. The method of claim 1, wherein generating a perspective
transform is based on the motion difference and image capture
parameters comprising one or more of a principal point, a focus
position, and a focal length.
7. The method of claim 1, wherein estimating the motion difference
of at least one row of the other rows includes interpolating
between a plurality of samples of motion data that are associated
with the at least one row to obtain motion information for the at
least one row.
8. The method of claim 1, wherein estimating the motion difference
of the row of the frame includes compensating for motion of the row
by finding an average of the motion of the row of the frame in the
current frame and motion of corresponding rows in one or more
adjacent images.
9. A method for reducing rolling shutter distortion in an input
frame, comprising: determining motion of each input row of the
input frame; finding a motion difference between the motion of a
base row of the input frame and each of the input rows; and
generating a plurality of transformed rows according to a plurality
of corresponding perspective transforms, wherein the perspective
transforms correct for shutter distortion associated with input
rows based on the motion difference.
10. The method of claim 9, further comprising assembling the
transformed rows to provide an output frame.
11. The method of claim 9, wherein the base row is a center row of
the input frame.
12. The method of claim 9, wherein the perspective transforms are
based on the motion difference and image capture parameters
comprising one or more of a principal point, a focus position, and
a focal length.
13. The method of claim 9, wherein determining motion of each input
row of the input frame includes interpolating between a plurality
of samples of motion data.
14. The method of claim 9, wherein determining motion includes
using motion data from a motion sensor.
15. A computer-readable storage device storing computer-executable
instructions that, when executed, cause a computer to execute a
method comprising: determining motion of each input row of an input
frame; finding a motion difference between the motion of a base row
of the input frame and each of the input rows; and generating a
plurality of transformed rows according to a plurality of
corresponding perspective transforms, wherein the perspective
transforms correct for shutter distortion associated with input
rows based on the motion difference.
16. An apparatus comprising: a image sensor to capture output a
sequence of input rows of an input frame; a motion sensor to output
motion data corresponding to motion of the apparatus at
approximately the time the image sensor captures the sequence of
input rows; a memory to store a corrected frame; and a controller
configured to: estimate motion of each input row based on the
motion data; find a motion difference for each row between the
motion of the row and a base row of the input frame; generate a
perspective transform for each row based on the motion difference
of the corresponding row; applying the corresponding generated
perspective transform to each row to generate a corrected row;
assembling the corrected rows into the corrected frame; and store
the corrected frame in the memory.
17. The apparatus of claim 16, wherein the base row is a center row
of the input frame.
18. The apparatus of claim 16, wherein the perspective transforms
are based on the motion difference and intrinsic camera parameters
comprising one or more of a principal point, a focus position, and
a focal length.
19. The apparatus of claim 16, wherein estimating motion of each
input row includes interpolating between a plurality of samples of
the motion data.
20. A chip comprising: a video device driver configured to receive
a captured frame including a plurality of captured rows; a motion
sensor driver configured to receive motion data associated with the
captured rows; and a rolling shutter distortion processor
configured to: determine motion of each captured row of the
captured frame; find a motion difference between the motion of a
base row of the captured frame and each of the captured rows; and
generating a plurality of transformed rows according to a plurality
of corresponding perspective transforms, wherein the perspective
transforms correct for shutter distortion associated with input
rows based on the motion difference.
21. The chip of claim 20, wherein the base row is a center row of
the captured frame.
22. The chip of claim 20, wherein the perspective transforms are
based on the motion difference and intrinsic camera parameters
comprising one or more of a principal point, a focus position, and
a focal length.
Description
PRIORITY CLAIM
[0001] The present application claims priority to U.S. Provisional
Application No. 61/657,706, filed on Jun. 8, 2012, and is a
continuation-in-part of U.S. patent application Ser. No.
13/209,899, filed on Aug. 15, 2011, the entirety of which are
incorporated by reference herein.
BACKGROUND
[0002] This disclosure relates generally to the field of image
processing. More particularly, but not by way of limitation, this
disclosure relates to compensating for unwanted image distortions
resulting from the so-called "rolling shutter" effect caused by
certain complementary-metal-oxide-semiconductor (CMOS) sensors
during video image capture operations.
[0003] Today, many personal electronic devices come equipped with
digital image sensors that are capable of capturing video composed
of a sequence of images. Exemplary personal electronic device of
this sort include, but are not limited to, mobile telephones,
personal digital assistants, portable music and video players and
portable computer systems such as laptop, notebook and tablet
computers. Many lower cost, high resolution cameras such as those
utilized in compact, portable personal electronic devices are
equipped with low-cost, low-power, CMOS sensors that can
potentially geometrically distort captured images if there is
movement of the device or the object being imaged while the CMOS
sensor is capturing the scene.
[0004] An image sensor converts photons into electrons, thus
converting optical images into electrical signals. Typically, image
sensors may be either a charge-coupled device (CCD) or a CMOS. A
CMOS sensor, unlike the CCD sensor, does not expose the entire
sensor array at the same time since it cannot store and hold all of
the individual pixel charges for the entire sensor array. Instead,
CMOS sensors employ a so-called "rolling shutter" technique,
wherein each row or scan line of the sensor array is exposed at
different times, read out sequentially (e.g., from the top of the
sensor to the bottom of the sensor), and then merged together to
form a single image.
[0005] As long as the camera device and the object being imaged are
stationary with each other, the output image typically does not
include any geometric distortions caused by the "rolling shutter."
However, if there is relative movement horizontally or vertically
between the image sensor and the object being imaged, the output
image may potentially be distorted or temporally sheared, as shown
in FIG. 1. This type of distortion is one example of what will be
referred to herein as the "rolling shutter effect." Resulting image
frames and video sequences suffering from rolling shutter
distortions are often aesthetically unpleasing and unwanted, as
they do not accurately represent the scene being captured. Further,
rolling shutter artifacts can worsen with high resolution images
and high frame rates, e.g., 1080p images captured at 30 frames per
second.
[0006] Some video capture devices now include "on board" motion
sensors, i.e., positional sensors (e.g., accelerometers and/or
gyrometers), which may be used to assist in various device
functions. For example, some devices may use gyrometer data to aid
in image stabilization by appropriately adjusting the device's lens
and/or sensor mechanism before an image or frame is captured. Once
captured, however, the image is retained as part of the video
sequence without substantial modification. This approach is not,
however, feasible for many devices incorporating video capture
capability. For example, at this time, it is generally considered
infeasible to provide movable lens mechanisms and the like in such
small form factor devices.
[0007] Accordingly, there is a need for techniques to reduce the
effects of rolling shutter distortion during image and video
capture in devices utilizing CMOS or other non-CCD image sensors.
By employing appropriate perspective transformations to captured
image data based on timestamped information gathered from
positional sensors in communication with the image capture device,
more efficient image processing techniques may be employed to
reduce the effects of rolling shutter distortion. By using novel
motion compensation techniques, informed by hardware motion
sensors, such as positional sensors, in communication with an image
capture device, a robust rolling shutter reduction system may be
employed, even in situations where reliably reducing rolling
shutter distortion effects was previously thought to be impossible
from either computational and/or power consumption standpoints.
BRIEF DESCRIPTION OF THE DRAWINGS
[0008] So that features of the present invention can be understood,
a number of drawings are described below. It is to be noted,
however, that the appended drawings illustrate only particular
embodiments of the invention and are therefore not to be considered
limiting of its scope, for the invention may encompass other
equally effective embodiments.
[0009] FIG. 1 shows exemplary rolling shutter distortions on the
display of an electronic image capture device.
[0010] FIG. 2 shows, in flowchart form, a rolling shutter
distortion reduction system, in accordance with one embodiment.
[0011] FIGS. 3A and 3B show, in block diagram form, two different
embodiments for correlating image data with motion data.
[0012] FIG. 4 shows, in flowchart form, motion data being processed
and attached to video data, in accordance with one embodiment.
[0013] FIG. 5 shows exemplary anchor rows and segments on the
display of an electronic image capture device and a timeline for
the read out of single image frame, in accordance with one
embodiment.
[0014] FIG. 6 shows, in flowchart form, a rolling shutter reduction
operation, in accordance with another embodiment.
[0015] FIGS. 7A and 7B illustrate specific aspects of a rolling
shutter reduction operation, in accordance with one embodiment.
[0016] FIG. 8 shows, in flowchart form, one technique to generate a
perspective transformation, in accordance with this disclosure.
[0017] FIG. 9 illustrates an exemplary method for a rolling shutter
reduction in accordance with an embodiment of the present
invention.
[0018] FIG. 10 illustrates exemplary rows on the display of an
electronic image capture device and a timeline for the read out of
a portion of an image frame, in accordance with one embodiment.
[0019] FIGS. 11A and 11B show, in a functional block diagram, two
illustrative devices capable of providing rolling shutter reduction
capability, in accordance with this disclosure.
[0020] FIG. 12 shows, in block diagram form, an electronic device,
in accordance with one embodiment.
DETAILED DESCRIPTION
[0021] The rolling shutter distortion reduction techniques
disclosed herein are designed to handle the processing of images
captured by handheld personal electronic devices having CMOS
sensors. More specifically, the techniques described herein provide
rolling shutter distortion reduction solutions by leveraging
timestamped positional sensor data received from positional sensors
in communication with an image capture device, wherein the
positional sensor information may be correlated with corresponding
captured image information that was read out from particular rows
of a CMOS sensor at substantially the same time that the
corresponding positional sensor information was recorded.
[0022] Rolling shutter reduction may be understood with reference
to any of the techniques disclosed in the inventors' co-pending
application, "Rolling Shutter Reduction Based on Motion Sensors,"
Ser. No. 13/209,899, filed Aug. 15, 2011, the disclosure of which
are incorporated herein in their entirety.
[0023] Generalized steps involved in rolling shutter reduction
techniques according to embodiments described herein include:
acquiring multiple motion samples from positional sensors in
communication with an image capture device per captured image
frame; defining a plurality of "anchor rows" for each captured
image frame, the anchor rows to be used as representatives of
device motion during the read out of particular portions of the
senor array; calculating a base motion for each of the plurality of
anchor rows based on the acquired motion samples corresponding to
the read out of the anchor row whose base motion is being
calculated; calculating a correction motion for each of the
plurality of anchor rows using multiple frame filtering;
calculating a 2D-perspective transform matrix for each of a
plurality of segments for each captured image frame, the matrices
based at least in part on an interpolation of the calculated
correction motion of the one or more anchor rows closest to the
segment whose transform matrix is being calculated; applying the
calculated perspective transform matrix to each segment of each
frame independently to generate corrected image segments;
assembling the plurality of corrected image segments for each image
into a corrected image; and storing the corrected image in a
memory.
[0024] According to an embodiment, generalized steps involved in
rolling shutter reduction techniques according to embodiments
described herein include: acquiring multiple motion samples from
positional sensors in communication with an image capture device
per captured image frame; defining a base row having a calculated
motion to which the detected motion in the rest of the frame may be
corrected; defining a plurality of anchor rows for the frame;
calculating a base motion for each of the plurality of anchor rows
based on the acquired motion samples corresponding to the readout
of the anchor row whose base motion is being calculated;
calculating a correction motion for each of the plurality of anchor
rows using the motion of the base row; calculating a 2D-perspective
transform matrix for each of a plurality of segments for the frame,
the matrices based at least in part on an interpolation of the
calculated correction motion of the one or more anchor rows closest
to the segment whose transform matrix is being calculated; applying
the calculated perspective transform matrix to each segment of each
frame independently to generate corrected image segments;
assembling the plurality of corrected image segments for each image
into a corrected image; and storing the corrected image in a
memory.
[0025] Thus, in one embodiment described herein, a rolling shutter
reduction method is disclosed comprising: obtaining, by a device
comprising an image sensor, a sequence of images, wherein each
image in the sequence comprises a plurality of sequentially read
out rows, and wherein the device has one or more image capture
parameters; selecting a first plurality of rows from a first image
in the sequence; for each of the first plurality of rows: obtaining
motion information corresponding to motion of the device at
approximately the time the row was obtained; and determining a
motion estimate for the row based, at least in part, on the
obtained motion information; defining a first plurality of
segments, the first plurality of segments comprising the first
image; for each of the first plurality of segments: generating a
perspective transformation for the segment based, at least in part,
on the motion estimates for one or more of the first plurality of
rows; and applying the generated perspective transformation to the
segment to generate a corrected image segment; assembling the
plurality of corrected image segments into a first corrected image;
and storing the first corrected image in a memory.
[0026] In another embodiment described herein, a rolling shutter
reduction method is disclosed comprising: obtaining a sequence of
images from an image sensor in a device, wherein each image in the
sequence comprises a plurality of rows, and wherein the device has
image capture parameters; selecting a first plurality of rows from
a first image in the sequence; determining a motion for each of the
first plurality of rows based on motion data from one or more
motion sensors of the device; determining a correction motion for
each of the first plurality of rows based, at least in part, on the
determined motion for the row; identifying a first plurality of
segments for the first image, wherein each segment comprises a
second plurality of rows; determining a correction motion for each
of the first plurality of segments based, at least in part, on the
determined correction motion for one or more of the first plurality
of rows; generating a transformation for each of the first
plurality of segments based, at least in part, on the segment's
determined correction motion; independently applying each of the
generated transformations to its corresponding segment to generate
a corrected image; and storing the corrected image in a memory.
[0027] In yet another embodiment described herein, an electronic
device, is disclosed comprising: an image sensor; a positional
sensor; a memory communicatively coupled to the image sensor; a
programmable control device communicatively coupled to the memory
and the positional sensor, wherein the memory has computer program
code stored thereon for causing the programmable control device to:
receive a plurality of sequential images captured by the image
sensor, the electronic device having image capture parameters,
wherein each of the plurality of sequential images is associated
with values corresponding to the image capture parameters at the
time each of the images was captured; obtain motion information
from the positional sensor for each of the plurality of sequential
images, wherein the motion information for each image in the
plurality of sequential images is obtained at approximately the
same time as each image was captured, and wherein the motion
information comprises a plurality of motion samples; divide each
image from the plurality of sequential images into a plurality of
segments; generate a perspective transformation for each segment of
each image based, at least in part, on one or more motion samples
corresponding to the segment and the one or more image capture
parameters associated with the image; apply the generated
perspective transformations independently to each segment of each
of the plurality of sequential images to substantially remove
rolling shutter distortions; and store each of the perspective
transformed plurality of sequential images in the memory.
[0028] Novel and improved image processing techniques for rolling
shutter distortion reduction, e.g., as may be used for handheld
personal electronic image capture devices having positional
information sensors, in accordance with the various embodiments
described herein may be implemented directly by a device's hardware
and/or software, thus making these robust rolling shutter reduction
techniques readily applicable to any number of electronic devices
with appropriate positional sensors and image processing
capabilities, such as mobile phones, personal data assistants
(PDAs), portable music players, digital cameras, as well as laptop
and tablet computer systems.
[0029] This disclosure pertains to devices, methods, and computer
readable media for reducing rolling shutter distortion effects in
captured video frames based on timestamped positional information
obtained from positional sensors (e.g., gyroscopic and
accelerometer sensors) in communication with an image capture
device. In general, rolling shutter reduction techniques are
described for generating and applying image segment-specific
transforms to already-captured segments (i.e., portions) of images
in a video sequence so as to counter or compensate for unwanted
distortions that occurred during the read out of the image sensor.
Such distortions may be due to, for example, the use of CMOS
sensors combined with the rapid movement of the image capture
device with respect to the scene being captured. In contrast to the
prior art, rolling shutter reduction techniques described herein
may be applied to captured images in real-time or near real-time
using positional sensor information, rather than in post-production
via processing-intensive image processing routines that would
require an analysis of the content of the underlying image
data.
[0030] The techniques disclosed herein are applicable to any number
of electronic devices with optical sensors and/or positional
sensors, such as digital cameras, digital video cameras, mobile
phones, personal data assistants (PDAs), portable music players, as
well as laptop and tablet computer systems.
[0031] In the interest of clarity, not all features of an actual
implementation are described in this specification. It will of
course be appreciated that in the development of any such actual
implementation (as in any development project), numerous decisions
must be made to achieve the developers' specific goals (e.g.,
compliance with system- and business-related constraints), and that
these goals will vary from one implementation to another. It will
be further appreciated that such development effort might be
complex and time-consuming, but would nevertheless be a routine
undertaking for those of ordinary skill having the benefit of this
disclosure.
[0032] In the following description, for purposes of explanation,
numerous specific details are set forth in order to provide a
thorough understanding of the inventive concept. As part of the
description, some structures and devices may be shown in block
diagram form in order to avoid obscuring the invention. Moreover,
the language used in this disclosure has been principally selected
for readability and instructional purposes, and may not have been
selected to delineate or circumscribe the inventive subject matter,
resort to the claims being necessary to determine such inventive
subject matter. Reference in the specification to "one embodiment"
or to "an embodiment" means that a particular feature, structure,
or characteristic described in connection with the embodiments is
included in at least one embodiment of the invention, and multiple
references to "one embodiment" or "an embodiment" should not be
understood as necessarily all referring to the same embodiment.
[0033] Referring to FIG. 1, an electronic image capture device 100a
is shown possessing a display and an exemplary CMOS sensor. As is
shown in FIG. 1, a black oval 108a is currently being displayed on
the device. As indicated by arrow 110, the CMOS sensor is "read
out" sequentially in descending rows 102 of pixels of the image
sensor. As indicated by arrow 112, the sensor in the example is
reading out from top to bottom chronologically, as measured in this
example in milliseconds. More specifically, it is this sequential
read out process that is often referred to in the art as a "rolling
shutter." The use of a rolling shutter means that there will be
different exposure times and different read out times for each row
on the image sensor. Typical image sensor packages are capable of
providing timestamps for the beginning of each frame. By combining
the knowledge of the timestamp for the beginning of a captured
frame with the knowledge of the read out speed of the image sensor,
the exact capture time for each particular row can be determined.
As will be discussed later, knowledge of the capture time for each
particular row may be important when attempting to correlate
timestamped positional sensor information with particular rows of
captured image data.
[0034] Turning now to electronic image capture device 100b, the
effects of moving the device 100b to the right while the black oval
108b that is being imaged remains stationary are illustrated.
Movement of device 100b is represented by arrow 104b. Specifically,
and as is further illustrated via the examination of vertical axis
106b, sections of black oval 108b that were captured later in time
chronologically by the CMOS sensor (i.e., those near the bottom of
the display on device 100b) will not align vertically with the
sections of black oval 108b that were captured earlier in time
chronologically by the CMOS sensor (i.e., those near the top of the
display on device 100b). This may result in the warped effect seen
on the display of device 100b, wherein the oval 108b appears to be
tilted to the right.
[0035] Likewise, turning to electronic image capture device 100c,
the effects of moving the device 100c to the left while the black
oval 108c that is being imaged remains stationary are illustrated.
Movement of device 100c is represented by arrow 104c. Specifically,
and as is further illustrated via the examination of vertical axis
106c, sections of black oval 108c that were captured later in time
chronologically by the CMOS sensor (i.e., those near the bottom of
the display on device 100c) will not align vertically with the
sections of black oval 108c that were captured earlier in time
chronologically by the CMOS sensor (i.e., those near the top of the
display on device 100b). This may result in the warped effect seen
on the display of device 100c, wherein the oval 108c appears to be
tilted to the left. Any time that an image capture device using a
CMOS sensor is moved quickly over a scene (sometimes referred to as
a "whip pan"), that is, the sensor is moved before there has been
enough time to read out the image data from the entire image
sensor, there is a potential for rolling shutter distortions to be
manifested in the captured images.
[0036] Referring to FIG. 2, rolling shutter reduction operation 200
in accordance with one embodiment begins by capturing a raw video
sequence 205 (block 210) and corresponding motion data 215 (block
220). Motion information 215 may then be attached to individual
frames within raw video sequence 205 (block 225) to produce video
sequence 230 with motion data. It can be advantageous to capture
motion data for each frame in raw video sequence 205 so that each
captured frame has a corresponding motion datum. It can also be
advantageous, and is common, for each frame in a video sequence
such as raw video sequence 205, to have a timestamp indicating when
the particular frame was captured and the read out speed of the
image sensor (e.g., during acts in accordance with block 210).
Frames within video sequence 230 may then be transformed, in
accordance with this disclosure, based on various motion estimates
made for multiple segments in each of the frames comprising video
sequence 230 to reduce the effects of rolling shutter distortions
(block 235). The result is a rolling shutter distortion reduced
video sequence 240 that may be written (block 245) to storage
250.
[0037] Referring to FIG. 3A, in one embodiment video capture
operation 210 may be performed by image sensor 300, and motion data
capture operation 220 may be performed by gyroscopic sensor
(gyrometer) 305. Image sensor 300 may provide black and white or
color images and use, for example, complementary metal-oxide
semiconductor (CMOS) technology. Gyro sensor 305 may be used to
generate rate data in three dimensions, e.g., (x, y, z) or (pitch,
roll, yaw), or in a quaternion system. Gyro sensor 305 may use any
desired technology such as micro-electromechanical systems (MEMS)
technology.
[0038] It will be understood that video captured in accordance with
block 210 (e.g., by image sensor 300) and motion data captured in
accordance with block 220 (e.g., by gyro sensor 305) should be
correlated. It is important that an image captured at time t0 be
synchronized with motion data captured at approximately the same
time. In the embodiment illustrated in FIG. 3A, image sensor 300
may signal gyro sensor 305 each time an image row is captured via,
for example, the Vsync and Hsync signals. Gyro sensor 305, in turn,
may tag each "next captured" motion datum each time a Vsync or
Hsync signal is received. This permits each frame in raw video
sequence 205 to be correlated or associated with the proper motion
data. Use of the phrase "next captured" reflects the possibility
that motion sensor 305 may operate on a different clock signal than
image sensor 300. That is, image sensor 300 and motion sensor 305
may operate asynchronously. Referring to FIG. 3B, in another
embodiment, common clock 310 may drive both image sensor 300 and
motion sensor 305. This arrangement permits the synchronous capture
of images and motion data. In another embodiment, common clock 310
may be used to generate timestamps for image sensor 300 and motion
sensor 305. In such an embodiment, the data acquisition of the
image sensor and the motion sensor are asynchronous, but the
timestamps are synchronized via common clock 310.
[0039] In some embodiments, the CMOS image sensor may be capturing
images at a rate of, e.g., 30 frames per second, while the
gyroscopic sensor 305 may be recording motion samples at a much
higher rate. For example, some gyroscopic sensors may sample at a
rate of 200 times per second. Thus, there may actually be a
plurality of motion samples, e.g., 6-7 motion samples, for each and
every image frame that is captured by the image sensor. As
mentioned above in reference to FIG. 1, CMOS sensors may be read
out sequentially, e.g., from the top of the sensor to the bottom of
the sensor. This read out process, although very rapid, does take
some finite amount of time. Thus, and as will be described herein,
for certain rolling shutter reduction techniques, it may be
important to correlate particular motion samples recorded by the
gyroscopic sensor 305 with certain representative rows of the image
sensor. These representative rows will be referred to herein as
"anchor rows." The number of anchor rows used in a particular
embodiment may range from a single anchor row all the way up to the
total number of rows in the image sensor (i.e., each row would
technically be an anchor row). Power consumption and processing
limitations may be important factors in dictating a suitable number
of anchor rows to be used for motion estimation within a captured
image. In one embodiment, six anchor rows, evenly spaced across the
breadth of the image sensor, are utilized for an image sensor
capturing 1,080 rows of pixels. As may be understood, choosing a
larger number of anchor rows may provide more granular estimations
of the motion of the image capture device during the capture of a
particular image frame. However, increasing the number of anchor
rows for which motion is estimated will come with the trade off of
requiring more power consumption and a greater number of
calculations to be performed by the image capture device.
[0040] Referring to FIG. 4, in one embodiment motion data 215 may
be attached to video data (raw video sequence 205) through a
process such as that illustrated in FIG. 4. First, it will be
understood that when a gyro such as sensor 305 is used to provide
motion data 215, what is actually produced is rate information: the
rate at which the video capture device is being moved in, for
example, each of 3 axes. Rate data may be integrated (block 400) to
produce instantaneous position information 405 (also in each of 3
axes). Using image timestamp information and motion detector tags
(which may also employ timestamps), each frame in raw video
sequence 205 may be associated with the appropriate position
information 405 (block 410). In another embodiment, operation 225
may also use accelerometer input 415 to assist in calibrating gyro
sensor 305's output and removing drifting. Also shown in FIG. 4 is
a high-level representation of a single image frame 420 from video
sequence 230. As shown, video frame 420 includes data 425
representing the image data itself (e.g., comprising a plurality of
image segments making up the image frame, wherein each image
segment comprises a plurality of rows of pixel data), a timestamp
of the first row 430, and the rolling shutter read out speed 435 of
the image sensor, which together provide sufficient information to
derive the times at which the various rows (e.g., the so-called
"anchor rows") of the image frame were read out from the image
sensor in accordance with block 210. After the attach operation
410, video frame 420 may also include an array of position
information 405 (i.e., motion data), wherein each positional
information sample corresponds to the position of the image capture
device at a particular point in time during the read out of the
various rows of the image sensor in accordance with block 220.
[0041] Referring to FIG. 5, exemplary anchor rows 500 and image
segments 502 are shown on the displays of electronic image capture
devices 100d and 100e, respectively, in accordance with one
embodiment. In one particular embodiment, and as is shown in FIG.
5, there are six anchor rows 500 corresponding to six different
read out rows from the image sensor of image capture device 100d.
As described above, by combining the knowledge of the timestamp for
the beginning of a captured frame with the knowledge of the read
out speed of the image sensor, the capture time 504 for each
particular anchor row can be determined with a fairly high degree
of precision. The capture times 504 for each of the anchor rows
(ROW1-ROW6) shown on the display of image capture device 100d are
listed to the left of image capture device 100d. In the exemplary
embodiment of 1080p video frames being captured at the rate of
thirty frames per second, it takes approximately 33.3 milliseconds
for the sensor to capture a single image frame. As such, the
following capture times are shown in FIG. 5 for illustrative
purposes: ROW1=1 ms; ROW2=8 ms; ROW3=15 ms; ROW4=21 ms; ROW5=27 ms;
ROW6=33 ms. It should be mentioned that, in the exemplary
embodiment shown in FIG. 5, each successively captured video frame
would have the same number of anchor rows placed at the same
locations over the video frame. In other words, anchor row ROW1 in
an exemplary video frame, e.g., video frame 15, in a sequence of
video frames would have a corresponding ROW1 in the previous video
frame, video frame 14, as well as the successive video frame, video
frame 16, as well as in all other video frames captured according
to the exemplary embodiment. As mentioned above, in one embodiment,
a gyroscopic sensor has a sampling rate of 200 Hz, meaning that it
reports a positional information sample readout every 5 ms.
[0042] Turning next to element 506 of FIG. 5, an exemplary timeline
for the read out of single image frame is shown. Across the top of
the timeline, time is listed out in one millisecond intervals from
0 ms to 35 ms. The first row of information in the timeline
corresponds to gyroscopic sensor read outs. As is shown, the
gyroscopic sensor used in the example of FIG. 5 samples at a rate
of 200 Hz (i.e., every 5 milliseconds), thus, the timeline shows
gyroscopic sensor samples g1-g8 occurring at: 0 ms, 5 ms, 10 ms, 15
ms, 20 ms, 25 ms, 30 ms, and 35 ms. As discussed above, the CMOS
sensor used in the example of FIG. 5 has read out the designated
anchor rows, ROW1-ROW6 at the following times: 1 ms, 8 ms, 15 ms,
21 ms, 27 ms, and 33 ms. For each anchor row, a device rotation
amount may be calculated by interpolating between the nearest
gyroscopic sensor samples, i.e., the one or more gyroscopic sensor
samples whose determined timestamps are closest to the determined
timestamp of the anchor row, as is shown by arrows 508 in FIG. 5.
For some anchor rows, such as ROW1, device rotation will be
calculated by interpolating between several gyroscopic sensor read
out samples, e.g., g1 and g2. For other anchor rows, such as ROW3,
there may be a single gyroscopic sensor read out sample, e.g., g4,
whose timestamp corresponds very closely with the read out time of
the anchor row, such that the device rotation may be calculated by
using the single gyroscopic sensor read out sample. As shown in
FIG. 5, both ROW3's readout and the recording of gyroscopic sensor
read out sample g4 occurred at hypothetical timestamp t=15 ms. Once
the capture time for each particular anchor row 500 is known, the
base motion for each anchor row 500 may be calculated based on the
interpolation of the recorded positional sensor information having
timestamps corresponding most closely to the timestamp of the
particular anchor row 500.
[0043] Referring now to FIG. 6, rolling shutter reduction operation
235 as implemented in one embodiment may begin once images making
up video sequence 230 begin to be received. Initially, the base
motion of each anchor row in an image frame may be characterized
with respect to corresponding anchor rows in a specified number of
"neighbor" frames (block 600). Referring to FIG. 7A, in one
embodiment the motion of an anchor row in a current frame (Fc)
captured at time td may be characterized by the corresponding
anchor rows of M number of previously captured frames (in this
example M equals three, i.e., the frames captured at prior times
ta, tb, and tc) and N number of subsequently captured frames (in
this example N also equals three, i.e., the frames captured at
later times te, tf, and tg). FIG. 7A plots the instantaneous
position of the corresponding anchor row of each of these frames
over time (represented as instantaneous motion signal 700). The
solid lines between successive points have been provided to
illustrate the "jittery" nature of motion data 215. It should be
understood that only a single axis of motion is represented in FIG.
7, but that, in many practical applications, motion in three
dimensions may be considered. It should also be noted that the
choice of three frames before and three frames after the current
frame is a design choice and may vary from implementation to
implementation depending on, for example, the image sensor (e.g.,
image sensor 300) and the particular type of video capture unit
being used (e.g., a professional stand-alone unit, a consumer
stand-alone unit, or embedded in a consumer device such as a mobile
telephone, portable music player or some other portable electronic
device). In other embodiments, M and N may both equal one. In still
other embodiments, e.g., in the case of an infinite impulse
response (IIR) filter, M may be the number of all previously
captured video frames.
[0044] Returning to FIG. 6, it is assumed that smooth motion in a
given direction is desired by the individual capturing the video
sequence. For example, the video capture device may be smoothly
panned to keep a specific target (e.g., a person) centered in the
frame. It follows that any jittery or high-frequency motion is
likely unintended (e.g., due to the individual's hand shaking).
With this as background, and the motion of each anchor row of a
frame characterized in accordance with block 600, the unwanted
aspects of the anchor rows' motion may now be estimated (block
605). Referring to FIG. 7B, to estimate the unwanted motion
components of the video capture device's movement, instantaneous
motion signal 700 may be filtered to eliminate its high-frequency
components (producing filtered motion signal 705). This may be
accomplished, for example, by passing instantaneous motion signal
700 through a low-pass filter or, more generally, an infinite
impulse response (IIR) or finite impulse response (FIR) filter. An
estimate of the unwanted motion for an anchor row at current frame
Fc (at time td) may then be given by the difference in the actual
position of the anchor row at frame Fc (at time td) and filtered
motion signal 705 (at time td) 710. In the example shown in FIGS.
7A and 7B, the current frame would be Fc (captured at time td) and
the prior frame is that frame captured at time tc. This process may
be repeated for the next "current" frame in a sliding-window
fashion. For example, in FIG. 7A the next frame to become the
"current" frame would be that frame captured at time te. Continuing
to use the three prior and three subsequent frame windows
introduced above, the prior frames upon which a new instantaneous
motion signal would be based are those frames captured at times td,
tc and tb. The successive frames upon which the new instantaneous
motion signal would be based are those frames captured at times tf,
tg and th (not shown).
[0045] Once an estimate of the unwanted motion for each anchor row
has been determined in accordance with block 605, that unwanted
motion may be subtracted from the motion of the anchor row that was
determined in block 600 to obtain the "correction motion," i.e.,
the calculated amount of corrective motion that may be applied in
order to reduce the effects of rolling shutter distortions, for the
anchor row. In some embodiments, such as is illustrated in FIG. 5,
once the correction motion has been calculated for each anchor row
500, the correction motion for each image segment 502 may be
interpolated 510 based on the calculated correction motion of the
nearest corresponding anchor rows. The correction motion along each
axis (e.g., x, y, z) may then be collected into a single 3.times.3
"correction motion rotation matrix." Hereinafter, the correction
motion matrix for each anchor row will be represented as rotation
matrix [R12], where the subscript `2` represents or identifies the
current frame and the subscript `1` represents or identifies a
prior frame. Following this, a 2D perspective transform matrix may
be calculated and applied independently to each row of the image
frame.
[0046] Ideally, a rolling shutter distortion reduction process
would be able to apply different correction motion estimates and
the resulting transformation matrices to each row of the image
frame. Such an implementation could be inefficient--especially when
implemented using a Graphics Processing Unit (GPU). In many
embodiments, the inventors have also discovered that applying a
separate transformation to each row is not necessary to achieve a
satisfactory reduction in rolling shutter distortion effects.
Rather than a separately calculated transformation matrix being
applied to each row, one embodiment described herein instead
applies one transformation matrix to each image segment 502
independently. An image segment is simply a consecutive group of
rows or columns from the captured image frame. The size of an image
segment is adjustable from 1 row all the way up to the total number
of rows captured by the image sensor array. In one embodiment, the
image segment height is chosen to be 32 rows. Thus, for a captured
image having 1,080 rows, there would be roughly 30 image segments.
Accordingly, roughly 30 separately calculated transformation
matrices would be applied to the corresponding image segments for
each frame. It may also be understood from the described
embodiments that each segment 502 may have one or more
corresponding motion samples recorded at approximately the time
that the consecutive rows comprising the segment were read out from
the image sensor.
[0047] Returning again to FIG. 6, once the correction motion for
each segment of the image frame has been determined, it may be used
to generate a perspective transformation (block 610) for each
segment. As discussed above, the size of image frame segments may
be predetermined in a given implementation depending on the power
and processing needs of the given implementation. Each segment's
perspective transformation may be independently applied to the
corresponding image segment in order to modify or compensate for
the unwanted motion caused by the rolling shutter effect (block
615). The result is rolling shutter reduced video sequence 240.
[0048] Referring now to FIG. 8, in one embodiment, perspective
transformation determination in accordance with block 610 begins by
obtaining various image capture device parameter values (block
800). Illustrative parameters include the focal length and focus
position used to capture a frame and the image capture device's
principal point. It will be recognized that on image capture
devices that provide the capability to move their lens and/or image
sensor assemblies, the focus position may change from frame to
frame. Based on the obtained parameters' values, the device's
intrinsic matrix may be found or generated (block 805). A
perspective transformation may then be determined for a particular
image segment using the image capture device's intrinsic matrix
associated with that frame (i.e., the intrinsic matrix generated
using device parameter values that were in place when the frame was
captured) and the image segment's "correction motion rotation
matrix" identified above (block 810).
[0049] A perspective transformation for a particular image segment
within a given frame may be derived as follows. First, it will be
recognized by those of skill in the art that the 2D projection of
real-space (which is 3D) onto an image sensor array (which is 2D)
may be given as--
( x y z ) = ( X Y Z ) , EQ . 1 ##EQU00001##
where
( X Y Z ) ##EQU00002##
represents a point in real-space, .PI. represents the image capture
device's intrinsic matrix and
( x y z ) ##EQU00003##
represents the 2D projection of the real-space point onto the image
sensor's plane. In essence, EQ. 1 represents a 3D-to-2D
transformation.
[0050] A novel use of this known relationship was to recognize
that--
( X Y Z ) = .PI. - 1 ( x y z ) , EQ . 2 ##EQU00004##
where
( x y z ) ##EQU00005##
represents a point in the sensor's 2D plane,
( X Y Z ) ##EQU00006##
represents an estimate of where that point is in real-space, and
.PI..sup.-1 represents the inverse of the image capture device's
intrinsic matrix described above with respect to EQ 1. Thus, EQ. 2
represents a 2D-to-3D transformation estimator.
[0051] Based on the discussion above regarding blocks 600, 605 and
FIG. 7, it will be recognized that--
( X 1 ' Y 1 ' Z 1 ' ) = [ R 01 ] ( X 1 Y 1 Z 1 ) , EQ . 3
##EQU00007##
where
( X 1 Y 1 Z 1 ) ##EQU00008##
represents the real-space location of a point at time t1,
[R.sub.01] the rotation matrix for frame-1 from time t0 (and frame
F0) to time t1 (and frame F1) mentioned before, and
( X 1 ' Y 1 ' Z 1 ' ) ##EQU00009##
represents the location of the same point after the estimated
unwanted motion has been removed.
[0052] From EQ. 2 we may obtain--
( X 1 Y 1 Z 1 ) = .PI. 1 - 1 ( x 1 y 1 z 1 ) , EQ . 4
##EQU00010##
where .PI..sup.-1 represents the inverse of the image capture
device's intrinsic matrix at time t1. Substituting EQ. 4 into EQ. 3
yields--
( X 1 ' Y 1 ' Z 1 ' ) = [ R 01 ] .PI. 1 - 1 ( x 1 y 1 z 1 ) . EQ .
5 ##EQU00011##
[0053] From EQ. 2 we may obtain--
( X 1 ' Y 1 ' Z 1 ' ) = .PI. 1 - 1 ( x 1 ' y 1 ' z 1 ' ) , EQ . 6
##EQU00012##
[0054] Substituting EQ. 6 into EQ. 5 yields--
.PI. 1 - 1 ( x 1 ' y 1 ' z 1 ' ) = [ R 01 ] .PI. 1 - 1 ( x 1 y 1 z
1 ) . EQ . 7 ##EQU00013##
[0055] Multiplying EQ. 7 by .PI..sub.1 yields--
.PI. 1 .PI. 1 - 1 ( x 1 ' y 1 ' z 1 ' ) = .PI. 1 [ R 01 ] .PI. 1 -
1 ( x 1 y 1 z 1 ) , EQ . 8 ##EQU00014##
which may be rewritten as--
( x 1 ' y 1 ' z 1 ' ) = .PI. 1 [ R 01 ] .PI. 1 - 1 ( x 1 y 1 z 1 )
. EQ . 9 ##EQU00015##
which may be rewritten as--
( x 1 ' y 1 ' z 1 ' ) = [ P 01 ] ( x 1 y 1 z 1 ) , EQ . 10
##EQU00016##
where [P.sub.01] represents the perspective transformation from
time t0 (and frame F0) to time t1 (and frame F1) for a particular
image segment within a given frame. Equations 9 and 10 describe how
to remove unwanted motion from rows comprising a particular image
segment at time t1 as reflected in rotation matrix [R.sub.01]. It
is also noted [P.sub.01] incorporates the image capture device's
parameters (e.g., focal length and focus position) at times t0 and
t1. More particularly, perspective transformation [P01] is based
solely on the image capture device's parameter values (e.g., focal
length and focus position) and determination of the image's
unwanted motion component. This information is available from
motion sensor 305 (e.g., a gyrometer). It will be recognized that
this information is computationally inexpensive to obtain and
process, allowing rolling shutter reduction operations in
accordance with this disclosure to be performed quickly and at low
computational cost.
[0056] According to an aspect of an exemplary embodiment, a
correction motion rotation matrix for a frame may be calculated
without reference to neighboring frames in the video sequence.
Therefore, this embodiment can be applied to a single image as well
as individual frames of a video sequence. FIG. 9 illustrates an
exemplary method of rolling shutter reduction operation 235, shown
in FIG. 2, according to an embodiment of the present invention.
[0057] The rolling shutter reduction operation 235 may include
determining the correction motion for a row by finding a difference
between the motion for that row and the motion of a base row (e.g.,
center row). The difference may be used to generate a perspective
transformation to be applied to the row to reduce the effects of
rolling shutter distortions. Thus, the rows of an image will have a
perspective transform that is applied to each of the rows to reduce
the effects of rolling shutter distortions with reference to the
base row.
[0058] Initially, the base motion of each frame may be
characterized with respect to a base row in the frame (block 900).
For a given image or video frame, a single base row may be
identified. For example, the base row may be identified as the
center row of the frame or another row of the frame. For example, a
row that coincides with a centroid of an object in the frame. The
base row may be one of the anchor rows.
[0059] The base row may be associated with motion data captured
from a motion sensor as described above. According to an
embodiment, the motion associated with the base row may be
estimated (block 900) by interpolating from the motion data of the
frame, the motion associated with the capture time of the base row
(e.g., the one or more gyroscopic sensor samples whose determined
timestamps are closest to the timestamp of the base row).
Similarly, the capture time of the base row may be estimated by
interpolating from known capture times of two rows in the frame.
The motion data associated with the base row may be considered the
base motion for the frame.
[0060] With reference to the base motion determined in accordance
with block 900, the unwanted aspects of the anchor rows' motion may
be estimated (block 905). An estimate of the unwanted motion for an
anchor row in the frame may be given by the difference in the
motion information for the anchor row from the base motion. The
motion information for the anchor row may be interpolated by
interpolating from the motion data of the frame, the motion
associated with the capture time of the anchor row. Additionally,
the capture time of the anchor row may be estimated by
interpolating from known capture times of two rows in the
frame.
[0061] The unwanted motion, the difference in the motion
information for the anchor row from the base motion, for each
anchor row may correspond to the "correction motion," i.e., the
calculated amount of corrective motion that may be applied in order
to reduce the effects of rolling shutter distortions, for the
anchor row. Although, obtaining the correction motion is discussed
with reference to the anchor row, the correction motion may be
determined for each row based on the motion data of each row and
the motion of the base row. The correction motion along each axis
(e.g., x, y, z) may then be collected into a single 3.times.3
"correction motion rotation matrix." Hereinafter, the correction
motion matrix for each anchor row (i) will be represented as
rotation matrix [R.sub.i]. Following this, a 2D perspective
transform matrix may be calculated and applied independently to
each row or anchor row of the image frame.
[0062] A perspective transformation for a particular image segment
within a given frame may be then be derived. In Equations 3-10,
[R.sub.01] may be replaced with [R.sub.i] where [R.sub.i] is the
rotation matrix for row i, and [P.sub.01] with [P.sub.i] such that
[P.sub.i] represents the perspective transformation from a
particular image segment within a given frame.
[0063] Once the correction motion for each anchor row or each
segment of the image frame has been determined, it may be used to
generate a perspective transformation (block 910) for each row or
segment as previously described with respect to FIG. 6. Each
segment's perspective transformation may be independently applied
to the corresponding image segment in order to modify or compensate
for the unwanted motion caused by the rolling shutter effect (block
915). The result is a rolling shutter reduced image or frame. For a
video sequence, each of blocks 900 through 915 may be executed for
each frame in the video sequence to generate a rolling shutter
reduced video sequence 240.
[0064] As discussed above with reference to FIG. 6, a perspective
transformation may be calculated for each row of an image frame.
The motion may be determined for each row by using the motion data
associated with each row to determine the difference in the motion
information for each of the rows from the base motion. For example,
FIG. 10 illustrates exemplary rows on the display of an electronic
image capture device and a timeline for the read out of a portion
of an image frame, in accordance with one embodiment.
[0065] Referring to FIG. 10, an image capture device 1000 may
include a display 1010 displaying an image having a plurality of
rows 1012. The rows 1012 may include anchor rows 1014 and a center
row 1016. In one particular embodiment, and as is shown in FIG. 10,
there are five anchor rows 1014 corresponding to five different
read out rows from the image sensor of image capture device 1000.
As described above, by combining the knowledge of the timestamp for
the beginning of a captured frame with the knowledge of the read
out speed of the image sensor, the capture time 1018 for each row
can be determined. In the exemplary embodiment of 1080p video
frames being captured at the rate of thirty frames per second, it
takes approximately 33.3 milliseconds for the sensor to capture a
single image frame. As mentioned above, in one embodiment, a
gyroscopic sensor has a sampling rate of 200 Hz, meaning that it
reports a positional information sample readout every 5 ms.
[0066] As shown in FIG. 10, across the top of the timeline, time
1018 is listed out in one millisecond intervals from 0 ms to 35 ms.
The first row of information in the timeline corresponds to
gyroscopic sensor read outs 1020. As is shown, the gyroscopic
sensor used in the example of FIG. 5 samples at a rate of 200 Hz
(i.e., every 5 milliseconds), thus, the timeline shows gyroscopic
sensor samples g1-g8 occurring at: 0 ms, 5 ms, 10 ms, 15 ms, 20 ms,
25 ms, 30 ms, and 35 ms. As discussed above, the CMOS sensor reads
out the rows of an image in subsequent rows (e.g., 1080 rows for a
1080p video frame). In FIG. 10, for clarity only rows 35, 260, 490,
540, 700, 880 and 1080 are illustrated. One or more of these rows
may correspond to an anchor row. For each row, a device rotation
amount may be calculated by interpolating between the nearest
gyroscopic sensor samples, i.e., the one or more gyroscopic sensor
samples whose determined timestamps are closest to the determined
timestamp of the anchor row.
[0067] For some rows, such as row 35, device rotation may be
calculated by interpolating between several gyroscopic sensor read
out samples, e.g., g1 and g2. For other rows, such as row 490,
there may be a single gyroscopic sensor read out sample, e.g., g4,
whose timestamp corresponds very closely with the read out time of
the anchor row, such that the device rotation may be calculated by
using the single gyroscopic sensor read out sample. The capture
time for each row may be provided as a timestamp or the capture
time may be determined by adding the capture time of the first row
and the readout time difference. Once the capture time for row is
known, the base motion for each row 1012 may be calculated based on
the interpolation of the recorded positional sensor information
1020 having timestamps corresponding most closely to the timestamp
of the particular row 1012.
[0068] The correction motion to be used for the perspective
transformation may be determined by determining the motion for a
base row (e.g., center row 540) and finding a difference in the
motion information for each row from the base motion. The
difference may be determined only for the anchor rows to reduce the
complexity of the computation. The correction motion along each
axis (e.g., x, y, z) may then be collected into a single 3.times.3
correction motion rotation matrix. Once the correction motion for
each row or each anchor row of the image frame has been determined,
it may be used to generate a perspective transformation for each
row or segment as previously described with respect to FIG. 6. Each
row's perspective transformation may modify or compensate for the
motion caused by the rolling shutter effect.
[0069] As discussed above with reference to FIG. 6, motion due to,
for example, individual's hand shaking, may be removed by comparing
the motion of one or more rows to the motion of corresponding rows
in frames captured before and/or after the current frame. The
removal of this motion may be performed before estimating and/or
removing the motion due to the rolling shutter effect or after
removing the motion due to the rolling shutter effect.
[0070] Referring to FIG. 11A, a functional view of illustrative
electronic device 1100 in accordance with this disclosure includes
video sensor 1105 (also referred to herein as a sensor array, or
image sensor), gyro sensor 1110, and accelerometer 1115. Video
sensor 1105 provides video frames to video device driver 1120, gyro
sensor 1110 provides motion data (e.g., rate of movement) to gyro
device driver 1125, and accelerometer 1115 provides its data to
accelerometer driver 1130. In the example of FIG. 11A, anchor rows
of the video frames and motion data are correlated through the use
of Vsync and Hsync signals as discussed above with respect to FIG.
3A. Gyro and accelerometer data may be collected to generate motion
data 1135 which may then be attached 1140 to the individual frames
within raw video sequence 205. Once motion data has been attached,
motion augmented video sequence 230 may be sent to rolling shutter
reduction processor 1145 which transforms each image segment of
each frame in accordance with its particular perspective
transformation to generate a rolling shutter reduced video sequence
240 that may then be written to storage 1150.
[0071] Referring to FIG. 11B, another illustrative video capture
device 1155 is shown. In this embodiment, however, common clock
1160 drives video 1105, gyro 1110 and accelerometer 1115 sensors.
As noted above with respect to FIG. 3B, use of common clock 1160
permits synchronous capture of image and motion data. In another
alternative embodiment (not shown), a common timer (or two distinct
timers driven by a common clock) may be used to add timestamps to
video frames and gyro samples. Specifically, video frames and gyro
samples may be generated with different clocks, but they may be
timestamped by a common clock, or two timers driven by a common
clock. In such an embodiment, the data acquisition may be
asynchronous, but the timestamps would be synchronized to a common
clock.
[0072] Referring now to FIG. 12, a simplified functional block
diagram of a representative electronic device possessing a display
1200 according to an illustrative embodiment, e.g., electronic
image capture device 100, is shown. The electronic device 1200 may
include a processor 1216, display 1220, proximity sensor/ambient
light sensor 1226, microphone 1206, audio/video codecs 1202,
speaker 1204, communications circuitry 1210, position sensors 1224
(e.g., accelerometers and/or gyrometers), image sensor with
associated camera hardware 1208, user interface 1218, memory 1212,
storage device 1214, and communications bus 1222. Processor 1216
may be any suitable programmable control device and may control the
operation of many functions, such as the generation and/or
processing of image metadata, as well as other functions performed
by electronic device 1200. Processor 1216 may drive display 1220
and may receive user inputs from the user interface 1218. Processor
1216 may be any suitable programmable control device or general or
special purpose processor or integrated circuit and may execute
instructions necessary to carry out or control the operation of
many functions, such as the generation and/or processing of image
metadata, as well as other functions performed by electronic device
1200. Processor 1216 may, for example, be a system-on-chip, such as
an applications processor found in a mobile device or a dedicated
GPU and may, for example, be based upon a RISC, CISC or any other
suitable architecture and may include one or more processing
cores.
[0073] Storage device 1214 may store media (e.g., image and video
files), software (e.g., for implementing various functions on
device 1200), preference information, device profile information,
and any other suitable data. Storage device 1214 may include one
more storage mediums for tangibly recording image data and program
instructions, including for example, a hard-drive, permanent memory
such as ROM, semi-permanent memory such as RAM, or cache. Program
instructions may comprise a software implementation encoded in any
desired language (e.g., C or C++).
[0074] Memory 1212 may include one or more different types of
memory which may be used for performing device functions. For
example, memory 1212 may include cache, ROM, and/or RAM.
Communications bus 1222 may provide a data transfer path for
transferring data to, from, or between at least storage device
1214, memory 1212, and processor 1216. User interface 1218 may
allow a user to interact with the electronic device 1200. For
example, the user input device 1218 can take a variety of forms,
such as a button, keypad, dial, a click wheel, or a touch
screen.
[0075] In one embodiment, the personal electronic device 1200 may
be an electronic device capable of processing and displaying media
such as image and video files. For example, the personal electronic
device 1200 may be a device such as such a mobile phone, personal
data assistant (PDA), portable music player, monitor, television,
laptop, desktop, and tablet computer, or other suitable personal
device.
[0076] Although the processes illustrated and described herein
include series of steps, it will be appreciated that the different
embodiments of the present disclosure are not limited by the
illustrated ordering of steps, as some steps may occur in different
orders, some concurrently with other steps apart from that shown
and described herein. In addition, not all illustrated steps may be
required to implement a methodology in accordance with the present
invention. Moreover, it will be appreciated that the processes may
be implemented in association with the apparatus and systems
illustrated and described herein as well as in association with
other systems not illustrated.
[0077] The foregoing description of preferred and other embodiments
is not intended to limit or restrict the scope or applicability of
the inventive concepts conceived of by the Applicant. As one
example, although the present disclosure focused on handheld
personal electronic image capture devices, it will be appreciated
that the teachings of the present disclosure can be applied to
other implementations, such as traditional digital cameras. In
exchange for disclosing the inventive concepts contained herein,
the Applicant desires all patent rights afforded by the appended
claims. Therefore, it is intended that the appended claims include
all modifications and alterations to the full extent that they come
within the scope of the following claims or the equivalents
thereof.
* * * * *